Bhattadev University Students: Collecting and Cleaning Your Own Data for an Analytics Internship

Yes, you can build a credible analytics portfolio piece from data you collect yourself, and a short internship is enough time if the question is small. Pick one narrow question, design a form with strict answer rules, collect 40 to 100 responses, clean them in a spreadsheet and document every decision in a validation rules sheet. This guide gives you the rules sheet, a 14-day plan and a worked illustrative example for a campus question.

Why primary data teaches more than a downloaded dataset

Downloaded datasets arrive pre-shaped. You never see how a vague question became a messy column. When you collect your own data, you face the problems analysts face daily: people type “Pathsala”, “pathsala ” and “Pathshala” for the same place; someone enters age as “twenty”; a skipped question leaves a blank you must interpret.

For an internship report, this is valuable. A reviewer can see that you defined a question, controlled how answers were captured, found problems and fixed them transparently. That chain is the core of entry-level analytics work.

It also keeps the scope honest. A 60-response survey will not support grand claims, but it can support careful ones, and learning to say that is part of the skill.

Choosing a question small enough to finish in 14 days

A good primary-data question has four properties: it concerns people you can reach, it can be answered in under two minutes, it has a decision attached, and it does not require sensitive personal information.

• Reachable: classmates, hostel residents, a neighbourhood market, a local library’s users.
• Short: 8 to 12 questions at most.
• Decision-linked: “Which study hours would a reading room need to open?” is better than “What do students think?”
• Low-risk: avoid health, caste, religion, income and exact addresses. You do not need them.

Examples of workable questions: how students commute to campus and what it costs; when students use the library or reading space; how households in one locality dispose of waste; which mobile-data problems affect online classes.

Consent, privacy and ethics for a student survey

Even a small survey needs ground rules. Put a short note at the top of the form stating who you are, the purpose, that participation is voluntary and that results will be shown only in aggregate. Do not collect names, phone numbers or email addresses unless the analysis truly needs them. If you need to prevent duplicates, use a respondent code you assign rather than identifying details.

Store the raw export in one place, keep it away from public sharing links, and do not publish row-level data in your report. If your question involves people outside your college, such as residents of a locality, tell them plainly what you are doing and accept a refusal without pressure.

Designing the form so the data arrives clean

Most cleaning work is created at the design stage. Every free-text box is a future cleaning job. Use these principles in Google Forms or any similar tool:

1. Prefer dropdowns and multiple choice for anything with a finite set of answers (department, year, mode of transport).
2. Use response validation for numbers: set a number field to accept only whole numbers between sensible limits.
3. Make only essential questions required. Forcing answers to sensitive items produces fake answers, which are worse than blanks.
4. Add an “Other (please specify)” option sparingly, and plan to recode those answers later.
5. Use a consistent scale, for example 1 to 5 where 1 is “very difficult” and 5 is “very easy”, and label both ends.
6. Ask one thing per question. “Is the library clean and quiet?” cannot be answered by someone who finds it clean but noisy.
7. Pilot with 5 people before launch and watch where they hesitate.

The validation rules sheet: a template you can copy

A validation rules sheet is a table that lists every field, what is allowed in it and what you will do when a value breaks the rule. Keep it as a separate tab in your workbook. It serves two purposes: it guides your cleaning and it becomes the documentation a reviewer reads. Below is a filled-in illustrative version for a campus commute survey.

The ranges above are illustrative choices for this example. Set yours from common sense about your population and write down why you chose them.

Cleaning in a spreadsheet: the order that avoids rework

Always work on a copy. Keep one tab called raw that you never edit, and a second called clean. Then follow this sequence:

1. Freeze and rename headers. Replace long question text with short snake_case names such as commute_minutes. Store the long wording in the rules sheet.
2. Trim whitespace. Use TRIM in a helper column so “Bus ” and “Bus” match.
3. Standardise case and spelling. Use PROPER or a lookup table that maps variants to one label.
4. Remove duplicates. Decide the rule first (same respondent code, or identical answers within a minute) and note how many rows were removed.
5. Fix data types. Convert numbers stored as text, and make dates real dates.
6. Apply range checks. Filter every numeric column for values outside the allowed limits and decide row by row.
7. Handle missing values. Leave blanks as blanks in the clean tab, and note the count per column. Avoid filling with averages for a small survey.
8. Recode “Other”. Move answers into existing categories where they clearly belong.
9. Create a change log. One row per action: what, how many rows, why.

Useful formulas: =TRIM(A2), =COUNTIF(F:F,">180"), =IF(ISNUMBER(G2),G2,""), and =COUNTBLANK(E2:E101) for missingness.

Worked illustrative example: a campus commute survey

This example is illustrative. The numbers are invented to show the method, and no analysis was actually run on them.

Question: How long and how costly is the daily trip to campus for students, and which mode feels hardest?

Imagined result: The form collects 72 responses. After export, the raw tab shows these issues:

• 6 rows are duplicates from people who submitted twice.
• 9 entries of commute mode read “auto”, “Auto ” or “shared auto”.
• 3 values of commute_minutes are 0, and 2 are above 180.
• 4 cost values contain symbols such as “₹300”.
• 11 respondents skipped the optional comment.

Cleaning decisions: Remove the 6 duplicates, leaving 66. Map all auto variants to “Shared auto”. Treat 0 minutes as blank because a trip cannot take zero minutes, and check the two long trips by asking the respondents if reachable; if not, mark them as outliers and report results with and without them. Strip currency symbols and convert to numbers.

Analysis the clean data allows: median commute minutes by mode; average monthly cost by mode; share of respondents rating difficulty 4 or 5. Report medians because a handful of long trips can distort averages.

Honest conclusion style: “Among 66 respondents reached through class groups, shared auto users reported the longest median trips.” Do not write “students at the university commute…” because your sample was not random.

A 14-day plan for collecting and cleaning your data

The free two-week online internship from CareerSwami runs for 14 days, so this plan fits that window. Adjust the daily load to your own schedule.

If responses are slow, extend collection by a day and shorten the writing time rather than lowering quality checks. Check data daily during collection so you can correct a confusing question early.

Sampling honestly: what a convenience sample can and cannot say

Sharing a form in class groups gives a convenience sample. That is acceptable for a learning project if you label it. State how you reached people, how many responded and who is probably missing, such as students without regular messaging access or those who rarely attend class. Compare your sample to what you know: if 80% of respondents are from one department, say so before interpreting anything.

Avoid percentages on very small groups. “2 of 5 hostel residents” reads better than “40%”.

Bhattadev University internships: where your data project fits

The university’s Internship Policy, effective from the 2023-24 session, makes internship mandatory for degree completion at 2 credits and says it can be done offline or online. It sets 60 hours for 2 credits and 120 hours for 4 credits, and asks for a completion certificate from the host agency showing hours of engagement. For students not evaluated by the external agency, the 2-credit internship is marked out of 50: report 20, presentation 20, viva 10.

That marking split matters for a data project. A report with a clear question, a documented rules sheet and a change log gives you material for the 20 report marks, and a clean summary tab gives you something concrete to present and defend in a viva. The policy also states that an Internship Cell oversees internships and that the placement officer sits on it.

Be careful about what is not established. Nothing verified here says whether a short online internship, including a 14-day one, would count towards the 60 hours, whether your department accepts a self-collected data project as the subject, or how a free certificate would be treated. The CareerSwami internship offers a free offer letter and a free completion certificate, but it should not be assumed to carry academic credit. The 2026 academic calendar lists a summer internship period, though the likely dates of 16 June to 31 July 2026 should be read from the original document. Affiliated college students should also look at the separate calendar for affiliated colleges.

Verification steps for Bhattadev students before you start

1. Download the Internship Policy PDF and highlight the sections on hours, host agency certificate and assessment.
2. Email your department coordinator or the Internship Cell asking in writing whether a 14-day online data analytics internship can count towards the required hours, and keep the reply.
3. Ask whether a self-designed survey project is acceptable as the report topic and whether a faculty guide must approve it first.
4. Confirm which calendar applies to you (university or affiliated colleges) and the exact summer internship dates, reading them from the original PDF.
5. Ask what report format, page limit and presentation format your department expects.
6. Check the eligibility condition on backlogs with your department office if it could affect you.

Documenting the project so a reviewer can trust it

A reviewer should be able to follow your work without asking you questions. Package these items:

• The question, population and sampling method in one paragraph.
• A screenshot or PDF of the final form.
• The validation rules sheet.
• The change log with counts: rows received, removed, recoded, blanked.
• A summary table and 2 or 3 charts, each with a one-line takeaway.
• A limitations paragraph.

Do not include respondent-level data in a public portfolio unless it has been fully anonymised and respondents agreed.

Frequently Asked Questions

How many responses do I need for a small analytics project?

For a learning project, 40 to 100 well-cleaned responses from a clearly defined group is workable. Quality of documentation matters more than size. State the number and who responded, and avoid strong claims about people you did not reach.

Should I use Google Forms or build the form directly in a spreadsheet?

Use a form tool for collection because it enforces answer types and timestamps automatically, then link or export responses to a spreadsheet. Typing answers directly into cells invites inconsistent entries, which means more cleaning.

What is the difference between a validation rule and a cleaning step?

A validation rule defines what a valid value looks like, such as a whole number from 1 to 180. A cleaning step is the action taken when data breaks the rule, such as flagging, recoding or blanking it. The rules sheet records both.

Is it acceptable to delete outliers?

Only with a reason. Remove values that are impossible, such as a trip of 0 minutes. For merely unusual values, keep them, flag them and show results with and without them so readers can see the effect.

How do I handle missing answers in a small survey?

Leave them blank in the clean tab, record how many are missing per question and calculate results on the answered rows. Filling blanks with averages can hide real uncertainty when the sample is small.

Can I collect data from people outside my college?

Yes, if you explain the purpose, keep participation voluntary and avoid personal identifiers. Choose a low-risk topic and say clearly in your report how those respondents were reached.

Will a self-collected data project satisfy my university’s internship report requirement?

That depends on your department. Ask your coordinator in writing whether a self-designed project is acceptable and whether a faculty guide must approve the topic, then follow the format they specify.

Can the CareerSwami internship be used to complete a data collection project like this?

The 14-day plan above is built to fit that free internship, and you can use the offer letter and completion certificate as evidence of your engagement. Whether your department accepts it for credit or hours is a separate question to settle with your coordinator.

Related Data Analytics guides

• UET Roorkee Students: Using Git and Version Control During a Data Analytics Internship
• Punjabi University Students: Statistics Refresher for Data Analytics Internship Tasks
• G.D. Goenka University Students: Asking Good Questions When Your Analytics Brief Is Vague
• All Data Analytics internship guides